An RR simulator was developed as part of the Computers in Cardiology Challenge (entry no. 184). The simulator was based on observed physiological changes as they exist in normal subjects.
Mean RR was dependent upon sleep and activity state. Mean RR was randomly chosen to lie within the range 0.6 to 1.2 s. During periods of activity mean RR fell between 0.1 and 0.2 s. The number of periods of activity during the wake time ranged from 3 to 7, with durations of 3 to 30 minutes. During the night the number of spontaneous arousals was between 5 to 25 with durations of between 5 and 20 s. Baseline changes were simulated by addition of low frequency sinusoids with periods of 5, 7 and 13 hours.
It is known that the frequency spectrum of normal RR shows a strong 1/f component and a distinct frequency peak at around 0.1 Hz. These were simulated by adding to the mean RR pink noise and a number of random phased sinusoids at around 0.1 Hz respectively. Pink noise was generated by filtering a white noise sequence.
Variation in RR due to respiration were dependent on sleep and activity states. The frequency of variation ranged from 0.06 to 0.3 Hz and the amplitude of variation from the mean RR ranged from 0.05 to 0.3 s.
A random number of sudden increases in RR to between 1.5 and 2 s were introduced during periods of activity to represent measurement artifacts.
We studied the characteristics of heart rate variability to enable simulators of beat-to-beat heart rate to be improved. Fifty sequences of beat-to-beat intervals covering periods of between 20 and 24 hours were studied. They were made available from PhysioNet. Information provided by PhysioNet indicated that approximately half of the interval sequences were from real recordings of normal subjects and the remaining from automated simulators.
The characteristics of the RR intervals were studied in both the time domain and frequency domain. Eleven characteristics were analysed, and the range of measurements for each was studied for outliers from the main distribution. For each characteristic measured, between 2 and 8 of the 50 sequences were classified as abnormal on this basis. If the distribution showed no values which could be classed as outliers, the extreme 4 were classed as abnormal. In the time domain, a restricted pattern of RR interval distributions classified 4 sequences as abnormal, and a reduced RR variability classified 10, with no overlap, giving a total of 14/50 as abnormal in the time domain. In the frequency domain, an abnormally restricted very low frequency pattern classified 17 as abnormal. The low frequency to high frequency ratio classified 4 as abnormal, but all these had already been detected by abnormal low frequency characteristics. Of the 17 classified in the frequency domain and of the 14 in the time domain there was an overlap of 9, resulting in 22 abnormal classifications, and suggesting that these were simulated. When this classification was assessed by PhysioNet a correct classification of 100% was achieved on a single entry (reference 20020426.082234). It was of note that the measurements for each of the characteristics assessed as abnormal tended to occur in pairs of strikingly similar values for the simulators, in agreement with the fact that two sequences for each simulator were included.
Simulators need to have improvements in the range of RR intervals, increased heart rate variability, and better very low frequency characteristics.
Supported by the National Institute of Biomedical Imaging and Bioengineering (NIBIB) under NIH grant number R01EB030362.
© PhysioNet Challenges. Website content licensed under the Creative Commons Attribution 4.0 International Public License.